Color Plate. Artist?s reconstruction of the new fossil species Hemignathus vorpalis (below), compared with 
adult males of H. wilsoni (Akiapolaau, above) and H. lucidus hanepepe (Kauai Nukupuu, middle). The fanciful 
plumage shown for the fossil species is based on the supposition that H. vorpalis resembled other members of 
its genus.
In 1992, well-preserved fossil skeletons 
of birds were discovered in a lava tube cave 
system in Pu`u Wa`awa`a (a traditional land 
district or ahupua`a), on the northern slope 
of Hualalai Volcano, island of Hawaii (Gi?   n 
1993). Subsequent explorations have revealed 
that several additional lava tube systems in 
the same district also contain avian skeletons. 
Fossils from the caves of the endangered Laysan 
Duck (Anas laysanensis), the endangered Nene 
(Branta sandvicensis), a large extinct fl ightless 
goose (Branta sp.), and four species of Hawaiian 
fi nches (Drepanidini) have been the subjects 
of research on ancient mitochondrial DNA 
(Cooper et al. 1996; Paxinos et al. 2002a, b) and 
comparative osteology (Livezey 1996, James 
2003). Other extinct birds found in the caves 
include fl ightless rails, a petrel, two crows, and 
a honeyeater (Meliphagidae).
Here, we describe a new species of large 
drepanidine bird with an extraordinary bill 
from Petrel Cave in the Pu`u Wa`awa`a district. 
The new species is known from a single partial 
associated skeleton. Even though the upper 
and lower bills are very di?  erent in length (see 
color plate), we are confi dent that the bones re-
ferred to the holotype are from one individual. 
They were found together by J. Gi?   n and R. 
Covington, isolated from other avian remains 
except for fi ve bones of an extinct fl ightless rail 
(Porzana sp.) and three unidentifi able bone frag-
ments (a femur shaft   and two vertebrae) that 
probably belong with the holotype. Gi?   n and 
Covington collected all visible bones, and three 
months later, H. F. James and Gi?   n returned to 
the site to search again. At that time, we also 
collected the thin, sparsely distributed layer of 
fi ne sediment from the vicinity of the collection 
site (~10 cm  
3
 of sediment was scraped up and 
sieved through 0.16 mm mesh screens), but our 
additional search produced only a few uniden-
tifi able bone fragments. Thus, we have investi-
gated and rejected the possibility that more than 
one individual of large passerine bird died near 
the collection site.
Abstract.?We describe a new species of drepanidine bird from a fossil found in a lava tube 
cave on the island of Hawaii. The bill of the new species combines a long, scimitar-like maxillary 
rostrum with a much shorter mandibular rostrum, giving the bird a most unusual appearance. 
The general bill form is shared with members of the drepanidine genus Hemignathus, to which 
the new species is referred, although it is quite distinct from other species of Hemignathus in as-
pects of its bill morphology and in its much larger body size. The scimitar-like maxilla appears 
to be adapted to probing in cracks or crevices and possibly to moving loose material, such as 
leaf detritus, to obtain invertebrate prey. The species became extinct <3,000 years ago. Received 
1 October 2000, accepted 20 April 2003.
Resumen.?Se describe una nueva especie de ave drep?nida sobre la base de un individuo 
f?sil recobrado en una cueva (tubo) de lava en la isla de Hawaii. El pico de la nueva especie 
es una combinaci?n de maxilar largo en forma de cimitarra con un rostro mandibular mucho 
m?s corto y recto, proporcion?ndole un aspecto muy particular al ave. Esta forma del pico 
es compartida con las especies del g?nero drep?nido Hemignathus, al cual es referida la 
nueva especie, que se diferencia de las dem?s por particularidades en la morfolog?a del 
pico y el tama?o mucho mayor del cuerpo. La semejanza de la porci?n rostral del maxilar 
a una cimitarra parece ser debida a una adaptaci?n para sondear en hendiduras o grietas, y 
posiblemente tambi?n para mover materia suelta, como son los detritos de hojas, en busca de 
invertebrados. La especie se extingui? hace menos de 3,000 a?os.
Department of Systematic Biology, National Museum of Natural History, Smithsonian Institution, 
P.O. Box 37012, Washington, D.C. 20013, USA
Helen F. James
1
 and Storrs L. Olson
A GIANT NEW SPECIES OF NUKUPUU 
(FRINGILLIDAE: DREPANIDINI: HEMIGNATHUS) 
FROM THE ISLAND OF HAWAII
The Auk 120(4):970?981, 2003
1
E-mail: james.helen@nmnh.si.edu
971
James and Olson
972
[Auk, Vol. 120
There is, unfortunately, a lack of consensus on 
the generic-level classifi cation of the drepanidine 
species known under the common names of nu-
kupuu, akiapolaau, akialoa, and amakihi. It is 
likely that molecular genetic and morphological 
studies currently underway will lead to a revised 
classifi cation. While we await those results, we 
hope to lessen confusion by using the same bino-
mial nomenclature as in our other recent papers 
(Olson and James 1995, James 2003). Thus, we 
restrict the genus Hemignathus to the species of 
drepanidines with long, thin maxillary rostra 
coupled with much shorter mandibular rostra. 
In addition to the new species, those taxa are the 
Nukupuu (Hemignathus lucidus), a polytypic spe-
cies known from four of the main islands from 
Kauai to Maui, and the Akiapolaau (H. wilsoni), 
a monoptypic species known from the island of 
Hawaii. Olson and James (1994) reported that a 
specimen of Hemignathus lucidus collected by the 
U.S. Exploring Expedition probably originated 
on the island of Hawaii, but R. C. Fleischer (pers. 
comm.) believes (on the basis of his analysis of 
mtDNA sequences from the specimen) that it 
originated on Oahu. We recognize the genus 
Akialoa for the species group known by the com-
mon name akialoa and formerly included in 
Hemignathus. Thus, we depart from the American 
Ornithologists? Union (AOU) Check-list of North 
American Birds (1998), which follows Pratt   
(1979) in defi ning the genus Hemignathus much 
more broadly. Those sources place the akialoas 
and the amakihis in Hemignathus, in addition 
to H. lucidus and H. wilsoni. The nomenclature 
used here is compared with that of AOU (1998; 
given in parentheses) as follows: Akialoa obscu-
rus (Hemignathus obscurus), Akialoa lanaiensis 
(Hemignathus ellisiana, lanaiensis group), Akialoa 
stejnegeri (Hemignathus ellisiana, stejnegeri group), 
Hemignathus lucidus (no change), Hemignathus 
wilsoni (Hemignathus munroi).
The new species would be allocated to 
Hemignathus under either of the two classifi ca-
tions. 
Comparative Material Examined
Modern skeletons.?Fringillidae: Drepanidini: 
Hemignathus wilsoni: MVZ 118830 female adult, 
MVZ 122610 male adult. Hemignathus lucidus: 
BMNH S/1961.11.39 unsexed adult. Akialoa ob-
scurus: BBM109 unsexed adult (skull and man-
dible removed from skin), BBM 4434 male adult 
(skull, mandible, and tarsometatarsus removed 
from skin). Akialoa stejnegeri: USNM 19094 un-
sexed adult.
Fossil specimens.?Akialoa sp., Hawaii (a fossil 
Akialoa that is much larger than Akialoa obscu-
rus of the same island, see James 2003): four 
partial skeletons from Umi`i Manu Cave, Pu`u 
Wa`awa`a, island of Hawaii, BPBM 173432-
179435. Akialoa cf. lanaiensis (see James 2003): 
fi ve maxillae and four mandibles from Pu`u 
Naio Cave, Maui, USNM 508652-508659 (some 
of them are small fragments); one mandible 
from Lua Lepo Cave, Maui, USNM 508661. 
Akialoa cf. stejnegeri: one maxilla and two man-
dibles from Maha`ulepu Cave, Kauai, USNM 
508662-508664. Akialoa upupirostris: a well-pre-
served skull with maxilla and fi ve partial man-
dibles from Maha`ulepu Cave, Kauai, USNM 
508665, 254171 (holotype), 255211, 255304, 
508666, and 508667. See Acknowledgments for 
explanation of museum acronyms.
Family Fringillidae
Tribe Drepanidini
Genus Hemignathus
Included species: Hemignathus lucidus 
(Lichtenstein, 1839), Hemignathus wilsoni 
(Rothschild, 1893), Hemignathus vorpalis new 
species.
Hemignathus vorpalis, sp. nov.
Holotype.?Partial associated skeleton, BPBM 
179437, consisting of a substantial portion of the 
maxilla, extending forward from the anterior-
most portion of the right nasal cavity toward 
the (missing) bill tip (Fig. 1A); the mandibular 
symphysis with the intermediate part of the left   
ramus and a portion of the intermediate part of 
the right ramus att  ached (Fig. 1B); a fragment 
of cranium involving the right auditory region 
(Fig. 1C); the proximal end and shaft   of the left   
humerus (Fig. 1D); the distal end and part of the 
shaft   of the left   ulna (Fig. 1E); the left   tibiotarsus 
lacking the distal end (Fig. 1F); the right tibiotar-
sus lacking both the distal end and most of the 
cnemial crests (Fig. 1G); the left   tarsometatarsus 
with the proximal end abraded (Fig. 1H); the 
proximal half of the right tarsometatarsus with 
the proximal end abraded (Fig. 1I). The piece of 
maxilla that is preserved has its dorsal surface 
abraded away along the posterior half. The 
New Species of NukupuuOctober 2003]
973
mandible, humerus, and right tibiotarsus have 
each been repaired from two or three pieces. 
The specimen was collected on 8 January 1993, 
by J. Gi?   n and R. Covington.
Type locality.?Petrel Cave, Hualalai Quad-
rangle, 155?51?W, 19?44?30?E (estimated to near-
est half minute), island of Hawaii, 1,200 m above 
sea level. The specimen was found near an open-
ing in the ceiling, through which we entered the 
lava tube with the assistance of climbing ropes. 
The cave was named by J. Gi?   n, in reference 
to the abundant bones of Dark-rumped Petrel 
(Pterodroma phaeopygia) found in it.
Distribution.?Known only from the type 
locality.
Age of holotype.?Holocene. Younger than 
Petrel Cave. Petrel Cave formed during a lava 
fl ow an estimated 3,000 to 1,500 years ago 
(Moore and Clague 1991).
Etymology.?The name vorpalis is from the 
adjective ?vorpal? as used by Lewis Carroll 
in the poem ?Jabberwocky? to modify the 
nouns ?sword? and ?blade? (?He took his 
vorpal sword in hand?. The vorpal blade 
went snicker-snack?), in reference to the long 
scimitar-like maxillary rostrum of the bird. 
Although no precise meaning has been given 
to ?vorpal? (Gardner 1960), there is an obvious 
connotation of potency associated with it in the 
poem. The specifi c epithet is formed by analogy 
with Latin carpus, carpalis, suggesting in turn 
that ?vorpal? might have been formed from a 
substantive ?vorpus,? the meaning of which is 
also obscure.
Measurements of holotype.?A number of the 
bones are broken in such a way that convention-
al measurements cannot be taken from them. 
Measurements such as length of fragment are 
Fig. 1. Photographs of the holotype (BPBM 197437) of Hemignathus vorpalis, new species. (A) Maxilla in lateral 
view. (B) Mandible in dorsal view. (C) Skull fragment in ventrolateral view. (D) Left humerus in anconal view. 
(E) Left tibiotarsus in anterior view. (F) Right tibiotarsus in anterior view. (G) Left tarsometatarsus in anterior 
view. (H) Right tarsometatarsus in anterior view. Scale bar = 2 cm.
James and Olson
974
[Auk, Vol. 120
given to identify the type specimen and for use 
in assessing size and shape. Maxilla: length of 
fragment, 30.0 mm; dimensions of the posterior-
most preserved cross-section?depth, 2.9 mm; 
greatest width, 1.7 mm; ventral width, 1.0 mm; 
dimensions of the anterior-most preserved 
cross-section?depth, 2.4 mm; greatest width, 
1.5 mm; ventral width, 0.3 mm. Mandible: 
symphysis length, 10.3 mm; width at angle 
of mandible, 6.9 mm (estimated); length from 
anterior margin of mandibular fenestra to tip 
of symphysis, 26.9 mm (includes a layer of en-
crustation of ~0.25 mm thickness); length from 
anterior margin of mandibular fenestra to pos-
terior margin of symphysis, 17.2 mm; length of 
intermediate part (lateral face) from the anterior 
margin of the mandibular fenestra to the distinct 
foramen at the angle of the mandible, 11.5 mm. 
Humerus: proximal width, 8.4 mm; greatest 
length from head of the humerus to distal angle 
of the deltopectoral crest, 9.6 mm; length of 
fragment, 22.4 mm. Ulna: distal width, 3.8 mm; 
length of fragment, 19.5 mm. Tibiotarsus: length 
from patellar crest to distal margin of the supra-
tendinal bridge, 48.3 mm; length from interar-
ticular area to distal articular face of the fi bular 
crest, 16.5 mm; proximal width including lateral 
cnemial crest, 8.8 mm. Tarsometatarsus: length 
Fig. 2. Comparison of the maxilla of Hemignathus vorpalis (BPBM 179437, holotype) with H. lucidus affinis 
(BMNH S/1961.11.39) and Akialoa sp. from Hawaii. (A) Skull of H. lucidus affinis in lateral view. (B, D, F) Maxilla 
of H. vorpalis in lateral, ventral, and dorsal views, respectively. (C, E, G) Maxilla of Akialoa sp. in lateral, ventral 
and dorsal views, respectively. Shaded portions of the maxilla of H. vorpalis are reconstructed using Akialoa sp. 
as a model. Sketches of the cross-section of the maxilla, in two places for each maxilla, are shown above the 
lateral views. Scale bar = 2 cm.
New Species of NukupuuOctober 2003]
975
from intercotylar eminence to the trochlea for 
metatarsal III, 37.6 mm; mid-shaft   width, 2.2 
mm; width across trochleae, 3.9 mm.
Diagnosis.?A passeriform bird, allied with 
the Passerida (Sibley and Ahlquist 1990) by the 
absence of pneumatic openings in the pneu-
motricipital fossa of the humerus, and with 
nine-primaried oscines by having a deep, but 
incompletely partitioned, pneumotri-
cipital fossa. The maxilla is long, 
decurved, and constricted anterior 
to the nasal openings (Fig. 2B, D), a 
combination of traits that it shares 
with the drepanidine genera Akialoa 
and Hemignathus. It resembles the 
genus Hemignathus as opposed to 
Akialoa in having the mandibular ros-
trum much shorter than the maxillary 
rostrum (Fig 3). However, the maxilla 
is much larger than in H. wilsoni or 
H. lucidus, and is more similar in size 
and general appearance to maxil-
lae of the species of Akialoa (Fig. 2). 
Compared with Akialoa, the maxilla is 
deep and narrow (compare sketches 
of cross-sections in Fig. 2) and is 
especially narrow on its ventral sur-
face (only 0.3 to 1.0 mm thick). The 
medial neurovascular sulcus is a very 
well-developed feature of the ventral 
surface of the maxilla in Akialoa and 
in the other species of Hemignathus 
(James 2003) but is lacking in H. vor-
palis. Instead, in H. vorpalis, neurovas-
cular tissue is transmitt  ed to the tip of 
the maxilla through an internal canal (see Fig. 2 
and description for further explanation).
The ventral profi le of the mandibular sym-
physis, in lateral view, is nearly straight, an 
aspect in which the new species resembles 
H. wilsoni, as opposed to H. lucidus and the 
various species of Akialoa, all of which have 
decurved symphyses (Fig. 4). In other respects, 
however, the mandible of H. vorpalis resembles 
H. lucidus and Akialoa more than it does H. wil-
soni. The symphyseal part of the mandible is 
not as short and stout as in H. wilsoni (Fig. 4). 
In H. wilsoni, the symphysis extends nearly as 
far posteriad as the tomial crests do; whereas 
in H. vorpalis, H. lucidus, and Akialoa, the sym-
physis does not extend posteriad nearly that 
far (Fig. 4). Also, H. wilsoni has the interme-
diate part of the mandibular ramus relatively 
short and deep compared with H. vorpalis, H. 
lucidus, and Akialoa (Fig. 4).
Compared with typical drepanidine tibio-
tarsi, that of H. vorpalis has a more pronounced 
crest for the ligamentum collaterale mediale.
The new species is inferred to be as large or 
larger in body size than any known drepani-
dine on the basis of dimensions of the post-
Fig. 3. Maxilla and mandible of Hemignathus vorpalis 
in lateral view, showing their inferred positions with 
respect to each other in the living bird. Shaded por-
tions of the maxilla of H. vorpalis are reconstructed 
using maxillae of Akialoa as models.
Fig. 4. Mandibles of Hemignathus vorpalis (BPBM 179437, ho-
lotype), H. lucidus affinis (BMNH S/1961.11.39), H. wilsoni (MVZ 
122610), and Akialoa sp. from Hawaii, in lateral and dorsal views. 
(A, E) H. vorpalis, new species. (B, F) Akialoa sp. (C, G) H. lucidus. 
(D, H) H. wilsoni. Scale bar = 2 cm.
James and Olson
976
[Auk, Vol. 120
cranial bones (see Comparative Osteometry, 
below, and Table 1).
Description.?The medial neurovascular sul-
cus on the ventral surface of the maxilla and the 
associated neurovascular tissue have not been 
lost but have become enclosed within the max-
illa, as shown in Figure 2. Cross-sections of the 
maxillae of H. lucidus and of a fossil Akialoa from 
Hawaii show a single canal enclosed within the 
maxilla and a distinct notch or indentation on 
the ventral surface indicating the position of 
the medial sulcus; whereas the cross-sections 
for H. vorpalis show two canals enclosed within 
the maxilla and no distinct notch on the ventral 
surface. The transformation of the medial neu-
rovascular sulcus into an internal canal in H. 
vorpalis apparently occurred in connection with 
narrowing of the ventral surface of the max-
illa. Along the anterior portion of the preserved 
fragment of the maxilla, the ventral surface con-
sists of litt  le more than two tomial crests meet-
ing along the median line. (The suture between 
the tomial crests is indicated by a median line 
in Figure 2D.) Along the posterior portion, the 
tomial crests do not meet but are separated only 
by a narrow space of bone.
In contrast with the maxilla, the mandible of 
H. vorpalis is not noticeably narrow compared 
with Akialoa or H. lucidus.
Remarks
The tip of the mandible of the holotype is cov-
ered by a thin fi ne-grained encrustation of min-
eral (Fig. 1). Aft  er mild acetic acid preparation 
failed to remove it, the mandible was X-rayed 
to examine the bone underneath the encrusta-
tion. The X-ray clearly showed that the tip of the 
mandible is unbroken and that the encrustation 
has a thickness of only ~0.25 mm.
Using the anterior rim of the nasal cav-
ity on the maxilla and the tomial crests of the 
mandible as landmarks, and taking a complete 
cranium and mandible of A. stejnegeri as an ana-
tomical analogue, we estimated the anatomical 
position of the maxilla over the mandible with 
the bill closed. That position is shown in Figure 
3 and was used to reconstruct the appearance 
of the bird in life (see color plate). The length 
of the maxilla also was estimated for the recon-
struction, by comparing the degree to which the 
depth of the maxilla tapers toward the bill tip in 
Akialoa with the degree to which it tapers in the 
fossil specimen. Although maxillae of Akialoa 
TABLE 1. Skeletal measurements (millimeters) of H. vorpalis compared with selected drepanidine taxa. The 
tibiotarsus length was taken from the interarticular area, to the medial condyle, thus omitting the patellar 
crest. The tibiotarsus width was taken from the lateral cnemial crest, to the rim of the medial articular face. 
The tarsometatarsus length was taken from the intercotylar eminence, to the trochlea for metatarsal III. It is 
not certain that the fossil finch, Orthiospiza howarthi, is correctly classified as drepanidine (H. F. James 
unpubl. data). 
Hemignathus Hemignathus Hemignathus Akialoa Rhodacanthis Orthiospiza 
 vorpalis wilsoni lucidus stejnegeri palmeri howarthi
Mandible:  
symphysis length 10.3 6.8, 7.7, 7.8  8.3 20.4 10.3 7.1 
Mandible: greatest width  
of symphyseal part 6.9
a
 5.7, 5.9 ? 5.9 12.5 11.1 
Humerus:  
length 27.1
a
 17.7, 18.0, 18.9 16.6 19.6, 19.8 24.5 25.0
b
Humerus:   
proximal width 8.4 5.5, 5.6, 5.7 5.7 6.1, 6.2 7.8 7.2 
Tibiotarsus:  
length 49.7
a
  30.5, 30.8 28.0 33.7, 34.0 ? 41.5 
Tibiotarsus:  
proximal width 8.8 5.5, 5.8 ? 5.4, 5.7 ? 6.7 
Tarsometatarsus:  
length 37.6 23.9, 24.4, 24.4  23.4  26.3, 26.5 27.5 ?
Tarsometatarsus:  
 width across trochleae 3.9  2.5, 2.6, 2.6  2.3 2.5, 2.6 3.0 ?
a
Dimensions of H. vorpalis that are estimated as explained in text. 
b
The humerus length of Orthiospiza howarthi is also estimated (see James and Olson 1991). 
New Species of NukupuuOctober 2003]
977
taper gradually toward the tip, the preserved 
portion of the maxilla of H. vorpalis tapers only 
slightly (Fig. 2), from which we surmise that the 
unbroken maxilla was considerably longer than 
the preserved portion, although we stress that 
the maxilla length shown in our reconstructions 
is conjectural. 
Phylogenetic Placement 
By their nature, fragmentary fossil birds pro-
vide limited clues to evolutionary relationships. 
In the case of H. vorpalis, the deeply excavated, 
but incompletely divided, nonpneumatic pneu-
motricipital fossa of the humerus suggests a 
relationship with the Drepanidini, the only 
group of native passerine birds in the Hawaiian 
Islands with that type of humerus. The very un-
usual bill form suggests a relationship with the 
genus Hemignathus.
James (2003) recently defi ned and scored 
osteological character variation in the drepani-
dine radiation for phylogenetic analysis. 
Hemignathus vorpalis was included in the study, 
but because the fossil was undescribed at the 
time, it was referred to as ?Hemignathus sp., 
Hawaii.? Parsimony analysis of the osteological 
data placed H. vorpalis in a clade with H. wilsoni 
and H. lucidus (fi g. 17 in James 2003). The genus 
Akialoa was depicted as a paraphyletic assem-
blage, basal to Hemignathus. Figure 5 shows 
the phylogenetic hypothesis for the Akialoa and 
Hemignathus clade from James? study, with the 
unambiguous character-state changes mapped. 
Notice that most of the character-state changes 
that contribute to the structure of the clade 
could not be observed in the fragmentary fossil 
of H. vorpalis. James? larger phylogenetic analy-
sis of the Drepanidini encompassed 61 taxa and 
65 informative characters, but the states for 43 
of those characters were recorded as missing in 
H. vorpalis. In view of the paucity of information 
about the new species, we consider its phyloge-
netic alliance with H. wilsoni and H. lucidus to 
be tentative. 
Comparative Osteometry
Skeletal comparisons of H. vorpalis with other 
Hemignathus and other large drepanidines (Table 
1) are limited because most of the bones of the ho-
lotype of H. vorpalis are either broken or missing. 
To make additional comparisons possible, certain 
dimensions of the broken bones were estimated 
using the methods described below.
The humerus length was estimated using the 
ratio of proximal width to length in unbroken 
drepanidine humeri. The ratio was determined 
for each of the six specimens of H. wilsoni, A. ste-
jnegeri, and Rhodacanthis palmeri listed in Table 
1 for which both proximal width and length of 
the humerus could be measured. Each ratio was 
used to estimate humerus length in H. vorpalis, 
yielding six estimates that ranged from 26.1 to 
28.4 mm, and averaged 27.1 mm. Similarly, the 
tibiotarsus length in H. vorpalis was estimated 
using the ratio of the length from the interartic-
ular area to the distal edge of the supratendinal 
bridge, to the total tibiotarsus length without 
the patellar crest. That ratio, determined in one 
specimen of H. wilsoni and two of A. stejnegeri, 
yielded estimates of tibiotarsus length in H. 
vorpalis of 49.5, 49.6, and 49.9 mm. Finally, the 
maximum width of the symphyseal part of 
the mandible was estimated visually under a 
dissecting microscope, using a caliper to ap-
proximate the correct measurement that would 
have been obtained if the bone were not slightly 
damaged in that region.
The dimensions of the postcranial skeleton 
of H. vorpalis exceed those of other species of 
Hemignathus. For example, the estimated hu-
merus length is 50% greater than the average for 
H. wilsoni and 63% greater than for H. lucidus. 
Table 1 also lists dimensions for the two other 
species of drepanidines with the largest postcra-
nial skeletons, R. palmeri and Orthiospiza howar-
thi (James and Olson 1991). (Orthiospiza howarthi 
may not be correctly classifi ed as drepanidine; 
see James 2003.) Like H. vorpalis, both species 
are extinct and only one partial postcranial skel-
eton is available for each. For the postcranial di-
mensions that can be compared among the three 
species, those of H. vorpalis are consistently the 
largest (8?31% larger than O. howarthi, 8?37% 
larger than R. palmeri.). Thus, with the caveat 
that the samples available for comparison are 
minimal, it appears that H. vorpalis is the largest 
species of drepanidine yet known.
Discussion
?The mere sight of so extraordinary a form 
could hardly fail to awaken in any one a keen 
desire to witness the manner of its feeding 
(Perkins 1893:106)
James and Olson
978
[Auk, Vol. 120
Although Perkins made the above comment 
about H. wilsoni, it would apply even more 
forcefully to H. vorpalis. Perkins was fortunate 
to witness and describe the foraging behavior 
of H. wilsoni, but in the case of H. vorpalis, we 
must be satisfi ed with what we can infer from 
morphology, phylogenetic context, and paleo-
ecology.
James? (2003) phylogenetic study places 
H. vorpalis within a clade of drepanidines 
(Hemignathus plus Akialoa) that favor bark and 
decaying wood as feeding substrates and feed 
primarily on insect larvae, spiders, and small 
beetles (Henshaw 1902, Perkins 1903, Munro 
1944, Ralph and Fancy 1996). Probing for ar-
thropods with the bill is a behavior common to 
all members of the clade whose foraging is well 
documented (Perkins 1893, 1903; Henshaw 1902; 
Munro 1944). Probing is done with the maxilla 
and mandible together in Akialoa and with the 
maxilla alone in Hemignathus. All of the species 
of Akialoa and Hemignathus that were observed 
in life by ornithologists were reported to feed 
on nectar at times (Perkins 1903, Munro 1944, 
Pratt   et al. 2001). Those species all possessed 
tubular brush-tipped tongues (Amadon 1950); 
and by phylogenetic inference, it is likely that 
H. vorpalis did as well.
Several accounts of the foraging behavior of 
H. wilsoni have been published (Perkins 1893, 
Fig. 5. Phylogenetic hypothesis for Akialoa and Hemignathus, from a study of osteology in the Drepanidini 
(James 2003). Brief descriptions of the character states that change unambiguously in the clade are given. 
Character numbers are from James (2003), which provides character state descriptions, the data matrix (matrix 
C), and phylogenetic analyses. Open bars identify characters that were not observable in the fragmentary fos-
sil of H. vorpalis; solid bars identify characters that were. The consistency index for each character is given in 
parentheses, based on parsimony analysis of 61 terminal taxa (see fig. 17 in James 2003).
New Species of NukupuuOctober 2003]
979
1903; Henshaw 1902; Munro 1944; Pratt   et al. 
2001). The bird creeps along trunks and branch-
es, inserting its thin, curved, somewhat fl exible 
upper bill into epiphytes, bark crevices, and 
burrows. It will grasp a piece of bark or lichen 
by wrapping the maxilla around it at one end, 
while bracing the chisel-like mandible against 
the other end, and then vigorously pull and pry 
until the bark or lichen comes loose, exposing 
any arthropods beneath. Hemignathus wilsoni 
can also chip away pieces of bark, wood, or li-
chen by hammering in a manner reminiscent of 
woodpeckers, apparently using the mandible 
alone, while employing its well-developed ca-
pacity for cranial kinesis to lift   the maxilla out of 
the way (see Munro 1944, Pratt   et al. 2001). Pratt   
et al. (2001) report that H. wilsoni sometimes 
uses its repertoire of hammering and prying 
with the mandible, probing and scooping with 
the maxilla, and wrenching with both bills to 
open stems and leguminous pods in search of 
larvae. They also report the recent discovery 
that H. wilsoni sometimes hammers on bark to 
open sap wells (Pratt   et al. 2001).
Hemignathus lucidus was reported to forage on 
trunks and branches in much the same manner 
as H. wilsoni, even making tapping noises with 
the bill (Perkins 1903). Likewise, A. obscurus 
and A. stejnegeri were reported to creep along 
branches and trunks, probing continually with 
the bill, and sometimes tapping and breaking 
o?   pieces of bark (Henshaw 1902, Perkins 1903). 
Akialoa also probes in leaf clusters of the `ie`ie 
vine (Freycinetia), cavities in tree fern (Cibotium) 
trunks, and similar places where debris collects 
and insects commonly shelter (Henshaw 1902, 
Perkins 1903). Hemignathus is reported to visit 
leaf clusters of the `ie`ie vine as well. The ac-
counts of Henshaw and Perkins suggest that 
there is considerable overlap in foraging niche 
among species in that clade, and indeed Perkins 
(1903) found the same arthropod species in the 
stomachs of the Hemignathus and Akialoa he col-
lected.
By phylogenetic inference, each of the previ-
ous foraging behaviors would be likely for H. 
vorpalis, but which would be possible for a bird 
with such a distinctive morphology? Certainly 
the maxilla is well constructed for probing in 
cracks and crevices. Tapping on bark could have 
been accomplished using either the maxilla or 
mandible by themselves. Forceful hammering 
may have been possible using the relatively 
straight mandible by itself, but that possibility 
is hard to evaluate because the posterior part 
of the mandible, which could be checked for 
adaptations for hammering (Zusi 1987), is not 
preserved. Nectar feeding would be feasible, 
although the bill appears not to be primarily 
adapted for nectarivory.
One behavior used frequently by H. wilsoni 
and H. lucidus would not have been possible for 
H. vorpalis. Assuming we have reconstructed 
its maxilla correctly, H. vorpalis could not have 
used its maxilla and mandible in forceful oppo-
sition to pull and pry up pieces of bark, because 
the maxilla is too long and not decurved enough 
to act in opposition to the mandible. In that re-
spect, the foraging behavior of H. vorpalis may 
have been more similar to that of Akialoa than to 
other species of Hemignathus.
The principal ways in which H. vorpalis di?  ers 
from other members of the clade are its large 
body size and its deep and narrow, relatively 
fl at-sided maxilla. Those may be specializations 
for foraging in a manner di?  erent from its rela-
tives. It is easy to envision the fl at-sided max-
illa being used not just for probing but also for 
creating openings in loose material such as leaf 
clusters, detritus, or even soil, by working the 
maxilla from side to side, for example. Perhaps 
H. vorpalis was more terrestrial than its relatives, 
using its long maxilla to move leaf litt  er in the 
manner of a Weka (Gallirallus australis) or kiwi 
(Apteryx spp.). Then again, H. vorpalis may have 
specialized in probing and fossicking in dense 
vegetation such as leaf clusters and epiphytes, 
or in any sort of receptacle where debris col-
lects in the forest vegetation. Unfortunately, 
the foraging behavior of this unusual bird will 
always be a matt  er for speculation rather than 
observation.
The habitat surrounding the entrance to 
Petrel Cave where the holotype of H. vorpalis 
was found is mesic montane forest, dominated 
by `ohi`a (Metrosideros polymorpha) and mamane 
(Sophora chrysophylla) trees. From what we know 
of the paleoecology of the site, there is no cause 
to reject the idea that H. vorpalis preferred forest 
habitat like its relatives. Tree molds in the sur-
rounding lava att  est to the presence of tall forest 
trees at the time of the lava eruption that formed 
the cave. In the vicinity of the type locality, those 
trees were killed by the same fl ow, sett  ing the lo-
cal vegetation back to an early stage of succes-
sion. Yet it is plausible that su?   cient time had 
James and Olson
980
[Auk, Vol. 120
elapsed for forest to return to the site before the 
individual of H. vorpalis died in the cave.
The extinction of H. vorpalis must have oc-
curred sometime aft  er Petrel Cave was formed 
by a lava fl ow, an estimated 3,000 to 1,500 years 
ago. No radiocarbon dates have been deter-
mined on bird bones from Petrel Cave, but four 
dates are available from bones of extinct birds 
found in Umi`i Manu, another lava tube in 
the same region that is estimated to be ~5,000 
years old (Moore and Clague 1991). The four 
dates range from 1,940 ? 50 (nsrl-2846) to 510 ? 
60 (beta-60548) radiocarbon years before pres-
ent (Paxinos et al. 2002a, b). In view of what is 
known about Holocene extinctions of birds in 
the Hawaiian Islands (e.g. Athens et al. 2002; 
Burney et al. 2001; James et al. 1987; Olson and 
James 1982, 1991; Paxinos 2002a, b), it is prob-
able that the extinction of H. vorpalis occurred 
sometime aft  er humans fi rst sett  led in the ar-
chipelago, 1,600 years ago or less. Excepting the 
endangered H. wilsoni, all the species in the ex-
traordinary genera Hemignathus and Akialoa ap-
pear to have become extinct in that time period 
(James and Olson 1991, Lepson and Johnston 
2000, Pratt   and Pyle 2000, Pratt   et al. 2001).
Acknowledgments
We thank J. Gi?   n, R. Covington, B. Schaefer, and 
B. Cobb for sharing the results of their cave explora-
tions with us, and A. Cooper for caving assistance 
in 1993. C. Kishinami and T. Lopez of the Bernice P.  
Bishop Museum, Honolulu, assisted with collecting, 
cataloging, and lending fossil specimens. We are 
grateful to J. Hume for his skillful reconstruction for 
the color plate and to T. Litwak (H. vorpalis) and J. 
Schulz (H. lucidus and H. wilsoni) for illustrations of 
osteology. At the USNM, we thank J. Jacobs for X-ray 
images, C. Hansen and J. Steiner for photography, 
P. Angle and F. Grady for helping to remove bones 
from rare skin specimens, and S. Suter for advice 
about phylogenetic analysis. Museum curators C. 
Cicero, N. Johnson, C. Kishinami, and R. Prys-Jones 
lent specimens in their care from the following in-
stitutions: BMNH (The Natural History Museum, 
Tring, United Kingdom), BPBM and BBM (Bernice P.  
Bishop Museum, Honolulu, Hawaii), MVZ (Museum 
of Vertebrate Zoology, Berkeley, California), USNM 
(National Museum of Natural History, Washington 
D.C.). Helpful comments on a previous version of 
the manuscript were provided by J. Gi?   n, T. Pratt  , 
and two anonymous reviewers. We thank W. Snar 
for spanish translation of the abstract. Field work 
in 1993 was supported by the Scholarly Studies 
Program of the Smithsonian Institution. Excavations 
in Maha`ulepu Cave in 1998, supported by National 
Science Foundation grant DEB9707260 to D. Burney, 
produced new fossils that were instrumental in as-
sessing the evolutionary relationships of H. vorpalis.
We dedicate this paper to the memory of Alan C. 
Ziegler.
Literature Cited
Amadon, D. 1950. The Hawaiian honeycreepers 
(Aves: Drepaniidae). Bulletin of the American 
Museum of Natural History 95:151?262.
American Ornithologists? Union. 1998. Check-
list of North American Birds, 7th ed. American 
Ornithologists? Union, Washington, D.C.
Athens, S. J., D. H. Tuggle, J. V. Ward, and D. J. 
Welch. 2002. Avifaunal extinctions, vegetation 
change, and Polynesian impacts in prehistoric 
Hawai`i. Archaeology in Oceania 37:57?78.
Burney, D. A., H. F. James, L. P. Burney, S. L. 
Olson, W. Kikuchi, W. L. Wagner, M. Burney, 
D. McCloskey, D. Kikuchi, F. V. Grady, R. 
Gage, and R. Nishek. 2001. Fossil evidence 
for a diverse biota from Kauai and its trans-
formation since human arrival. Ecological 
Monographs 71:615?641.
Cooper, A. J., J. Rhymer, H. F. James, S. L. 
Olson, M. Sorenson, R. C. Fleischer, and C. 
McIntosh. 1996. Ancient DNA and island en-
demics. Nature 381:484.
Gardner, M. 1960. The Annotated Alice. Penguin 
Books, London.
Giffin, J. 1993. New species of fl ightless birds 
found at Pu`u Wa`awa`a. `Elepaio 53:1?3.
Henshaw, H. W. 1902. Birds of the Hawaiian 
Islands, Being a Complete List of the Birds 
of the Hawaiian Possessions with Notes on 
Their Habits. Thomas G. Thrum, Honolulu, 
Hawaii.
James, H. F. 2003. The osteology and phylogeny 
of the Hawaiian fi nch radiation (Fringillidae: 
Drepanidini), including extinct taxa. Zoological 
Journal of the Linnean Society 135: in press
James, H. F., and S. L. Olson. 1991. Descriptions 
of Thirty-two New Species of Hawaiian 
Birds. Part II. Passeriformes. Ornithological 
Monographs, no. 46.
James, H. F., T. W. Stafford, Jr., S. L. Olson, D. W. 
Steadman, P. S. Martin, and P.  McCoy. 1987. 
Radiocarbon dates on bones of extinct birds 
from Hawaii. Proceedings of the National 
Academy of Sciences USA 84:2350?2354.
Lepson, J. K., and S. M. Johnston. 2000. Greater 
`Akialoa (Hemitgathus ellisianus) and Lesser 
`Akialoa (Hemignathus obscurus). In The Birds 
of North America, no. 512 (A. Poole and F. 
Gill, Eds.). The Birds of North America, Inc., 
Philadelphia.
New Species of NukupuuOctober 2003]
981
Livezey, B. C. 1996. A phylogenetic analysis of 
geese and swans (Anseriformes: Anserinae), 
including selected fossil species. Systematic 
Biology 45:415?450.
Moore, R. B., and D. A. Clague. 1991. Geologic 
map of Hualalai Volcano, Hawai`i. U.S. 
Department of the Interior, Miscellaneous 
Investigations Series, Map I-2213 (sheets 
1 and 2). United States Geological Survey, 
Washington, D.C.
Munro, G. C. 1944. Birds of Hawaii. Tongg 
Publishing Company, Honolulu, Hawaii.
Olson, S. L., and H. F. James. 1982. Prodromus of 
the Fossil Avifauna of the Hawaiian Islands. 
Smithsonian Contributions to Zoology, no. 365.
Olson, S. L., and H. F. James. 1991. Descriptions 
of Thirty-two New Species of Hawaiian Birds. 
Part I. Non-Passeriformes. Ornithological 
Monographs, no. 45.
Olson, S. L., and H. F. James. 1994. A specimen of 
Nuku pu`u (Aves: Drepanidini: Hemignathus 
lucidus) from the island of Hawai`i. Pacifi c 
Science 48:331?338.
Olson, S. L., and H. F. James. 1995. Nomenclature 
of the Hawaiian Akialoas and Nukupuus 
(Aves: Drepanidini). Proceedings of the 
Biological Society of Washington 108:373?387.
Paxinos, E. E., H. F. James, S. L. Olson, J. Ballou, 
J. A. Leonard, and R. C. Fleischer. 2002a. 
Prehistoric decline of genetic diversity in the 
Nene. Science 296:1827.
Paxinos, E. E., H. F. James, M. D. Sorenson, S. L. 
Olson, J. Jackson, and R. C. Fleischer. 2002b. 
MtDNA from fossils reveals a radiation of 
Hawaiian geese recently derived from the 
Canada Goose. Proceedings of the National 
Academy of Sciences USA 99:1399?1404. 
Perkins, R. C. L. 1893. Notes on collecting in Kona, 
Hawaii. Ibis 5:101?112.
Perkins, R. C. L. 1903. Vertebrata. Pages 365?466 
in Fauna Hawaiiensis, vol. 1, no. 4 (D. Sharp, 
Ed.). Cambridge University Press, Cambridge, 
United Kingdom.
Pratt, H. D. 1979. A systematic analysis of the en-
demic avifauna of the Hawaiian Islands. Ph.D. 
dissertation, Louisiana State University, Baton 
Rouge.
Pratt, T. K., and R. L. Pyle. 2000. Nukupu`u in 
the twentieth century: Endangered species or 
phantom presence? `Elepaio 60:35?41.
Pratt, T. K., S. G. Fancy, and C. J. Ralph. 2001. 
`Akiapola`au (Hemignathus munroi) and 
Nukupu`u (Hemignathus lucidus). In The Birds 
of North America, no. 600 (A. Poole and F. 
Gill, Eds.). The Birds of North America, Inc., 
Philadelphia.
Ralph, C. J., and S. G. Fancy. 1996. Aspects of the 
life history and foraging ecology of the endan-
gered Akiapolaau. Condor 98:312?321.
Sibley, C. G., and J. E. Ahlquist. 1990. Phylogeny 
and Classifi cation of Birds: A Study in 
Molecular Evolution. Yale University Press, 
New Haven, Connecticut.
Zusi, R. L. 1987. A feeding adaptation of the jaw 
articulation in New World jays (Corvidae). 
Auk 104:665?680.
Associate Editor: R. Zink